Claude Video 2.0 skill

A Claude Code skill that builds motion graphics in code: 21 ready looks, five questions, three approval stills and an MP4 render.

  • 143 ★ on the repository
  • Last commit
Skill

Extended access

Why this level

  • Scripts run Node, ffmpeg and a render on your machine.
  • Higgsfield AI clips spend credits under the key from `.env`.
All reasons and checks

itsssssjack/claude-video-2

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/claude-video-2-0

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Scripts run Node, ffmpeg and a render on your machine.
  • Higgsfield AI clips spend credits under the key from `.env`.

This is third-party code. Review the repository files before installing.

What it does

The skill walks Claude Code from picking a style to a finished MP4. A gallery of 21 styles opens first, and every tile plays a real video. Claude then asks five short questions: purpose and length, on-screen text, looping, brand colours and fonts, and your own media. The agent then writes a storyboard, shows three stills for approval and renders the piece with HyperFrames and GSAP. After the render it checks pacing and, for loops, compares the first and last frames pixel by pixel. Your own photos and clips or AI clips from Higgsfield can go into the video: before generating, the agent shows a plan with the credit cost and waits for a yes.

Who it is for. Creators of YouTube intros, Shorts and launch teasers who want studio-grade motion graphics without hand editing.

Good fit when

  • You need an intro for YouTube, a Short or a launch teaser.
  • The idea reads best as a diagram or a data story, for example in the `explainer-diagram` and `data-story` looks.
  • The animation loops and the last frame must match the first.
  • Your own photos and clips go into the video and the styling has to stay in one look.

Not a fit when

  • You need a long edit of live footage with many tracks.
  • Node and ffmpeg cannot be installed.
  • You want AI clips without spending Higgsfield credits.

Example request

Make me a motion graphic: a 12-second launch teaser in the swiss-kinetic look, 16:9, no loop.

Limitations

Node 22 or newer and ffmpeg are required. The skill checks for them and suggests the command to install anything missing. AI clips go through Higgsfield: the `HIGGSFIELD_API_KEY` key sits in the project `.env` file and generation spends credits. The Switzer font loads from Fontshare at render time, so a build needs network access. Installation is manual: the folder is copied into `~/.claude/skills/`. The author states that the skill works on the regular Claude Pro plan.

How to disable. Delete the `~/.claude/skills/claude-video-studio` folder. Remove the Higgsfield key from the project `.env` separately.

FAQ

Which looks does the gallery have?

Twenty-one: 13 studio pieces, among them `paper-world`, `ink-line` and `data-story`, and 8 code styles, among them `swiss-grid` and `risograph`. Open the gallery with `python3 ~/.claude/skills/claude-video-studio/scripts/gallery.py`.

Is a paid key required?

Only for Higgsfield AI clips. Pure graphics and videos built from your own photos and footage need no key.

What does the skill ask before building?

Five questions in one message, each with a default: purpose and length (12 seconds, 16:9), text, looping (no), brand style and media (none). The list lives in `questions.md` and can be edited.

How does the skill check the result?

From real frames. The user approves three stills, then a pacing check runs, and for loops the pixels of the first and last frames are compared.

Can I use my own photos and clips?

Yes, the fifth question is about media: your own photos and video, Higgsfield AI clips or nothing. In the gallery a studio piece can be flipped to the With media variant to see the same design with real footage.

Discussion

No questions or reviews yet.

Skills
Editors’ pick

A video editing skill for Claude Code and Codex: the agent cuts pauses and filler words, grades color, burns subtitles and adds animated overlays

SkillExtended access30.6KRepository stars

The video-explainer skill turns a script into a vertical 9:16 explainer: narration, captions, Remotion and a pre-render review

SkillExtended access122Repository stars

Claude Code skills that turn a talking-head recording into a YouTube video with graphics, sound and thumbnails

SkillBroad access368Repository stars

Five Claude Code skills that produce faceless vertical YouTube Shorts from Remotion code, video models or paper collage

SkillBroad access309Repository stars

Fifteen Claude Code skills for launch films, animated charts and reels made entirely from code, with no After Effects

SkillExtended access573Repository stars

A Claude Code skill that records a product in the browser, composes a video with chapters and captions and masters the MP4 locally

SkillExtended access373Repository stars

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialRepository starsMain difference
Claude Video 2.0 (this card)no143
HyperFramesno79,556gives the agent the HyperFrames engine itself; Claude Video 2.0 adds a style gallery, five questions and frame checks
Video Studio Skillno373records a product in the browser and assembles a chaptered video; Claude Video 2.0 draws graphics in code from a ready style

README in short

The author builds the skill on five habits: one style for the whole piece, real references instead of adjectives, a brief in numbers, one clock for loops and checking frames instead of code. Each habit is explained in `references/hacks.md`.

README badge

Original SKILL.md

---
name: claude-video-studio
description: Claude Video 2.0. Make studio-quality motion graphics with Claude Code, free. It opens a gallery of 21 looks playing as real videos so the person picks one, asks 5 short questions, then builds the piece in code (HTML + GSAP, rendered to MP4 with HyperFrames). Pure graphics, or with the person's own photos and clips, or with AI clips from Higgsfield behind a storyboard and credit check that waits for "yes". Every piece is checked from real frames: 3 stills to approve, a pacing check, and a pixel loop check. Use when someone says "make me a motion graphic", "make a video intro", "animate this", "make a loop", "make a title card", "motion graphics with my photos", "turn my photo into a clip", "Claude Video", or "/claude-video-studio".
---

# Claude Video 2.0

Three stages: **A. pick a look** (gallery) -> **B. five questions** -> **C. build** (storyboard, stills, render, checks).
Every step follows the 5 hacks in `references/hacks.md`: one style, real references, a brief in numbers,
one loop clock, and checking frames, not code.

`$SKILL` below means this skill's folder (usually `~/.claude/skills/claude-video-studio`).
Make a new work folder for each piece (for example `./video-<short-name>/`). Never build inside `$SKILL`.

## Before you start

Run `bash $SKILL/scripts/doctor.sh` once. If something is missing, give the person the one fix command and stop.

## Stage A: pick a look

1. Start the gallery in the background (use your background-run option; it serves on 127.0.0.1 and
   stops by itself after 90 minutes):
   `python3 $SKILL/scripts/gallery.py`
   It opens in their browser and prints the address. Give them the address too, in case it did not open
   (add `--no-open` if you only want the address). Stop it once they have picked, or leave it: it stops by itself.
2. Say, in one short message:
   "Your style gallery is open: <address>. Every tile is a real video made from code. Flip a tile to
   'With media' to see it with real photos and footage. Press 'Use this style' on the one you like and
   paste it here. Or paste 1 to 3 links or images of your own reference."
3. When they answer:
   - A style id (like `paper-world`): read `$SKILL/styles/<id>/STYLE.md`. That spec is now law for this piece.
   - Their own references: look at each one (open links, read images). Write a new spec in the work folder,
     `STYLE.md`, in the format of `$SKILL/references/style-format.md`. Describe HOW each reference was made
     (masked text, slow camera, long holds), with hex colours and easing numbers. Show it in 5 lines.
   - "You pick": choose the best look for their topic and say why in one line.

## Stage B: the questions

Read `$SKILL/questions.md`. Send the block between its two `---` lines as ONE message, word for word.
Skip anything they already told you. Accept "go" as yes to every default. Use its follow-ups only when needed.

## Stage C: build

1. **Brief in numbers** (hack 3). Write `BRIEF.md` in the work folder: size (1920 x 1080 for 16:9,
   1080 x 1920 for 9:16), length, 30 fps, loop or not, the exact words, palette in hex (brand colours replace the

Report a problem

Foxx AIClaude Video 2.0
Foxx AI

I am Foxx AI and I have already vetted this tool. Ask about install, setup or anything else, and I will keep it simple.